Culture Compliance and Confidentiality Study of Taxpayer Behavior in The United States

National Tax Association
Culture, Compliance, and Confidentiality

Author(s): James Alm, Michele Bernasconi, Susan Laury, Daniel J. Lee and Sally Wallace
Source: Proceedings. Annual Conference on Taxation and Minutes of the Annual
Meeting of the National Tax Association , Vol. 109 (2016), pp. 1-38
Published by: National Tax Association
Stable URL: https://www.jstor.org/stable/10.2307/26816608
REFERENCES
Linked references are available on JSTOR for this article:
https://www.jstor.org/stable/10.2307/26816608?seq=1&cid=pdf-
reference#references_tab_contents
You may need to log in to JSTOR to access the linked references.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms
National Tax Association is collaborating with JSTOR to digitize, preserve and extend access to
Proceedings. Annual Conference on Taxation and Minutes of the Annual Meeting of the
National Tax Association
This content downloaded from

103.4.95.4 on Mon, 17 Jul 2023 05:02:30 +00:00
All use subject to https://about.jstor.org/terms
February 2016
Culture, Compliance, and Confidentiality:

A Study of Taxpayer Behavior in the United States and Italy
James Alm
Michele Bernasconi
Susan Laury
Daniel J. Lee
Sally Wallace*
Abstract:
This paper analyzes the impact of confidentiality of taxpayer information on the level of
compliance in two countries with very different levels of citizen trust in government – the United
States and Italy. In both countries, the payment of the individual income tax relies heavily on
voluntary compliance, in which individuals are promised that information will be kept
confidential, at least until evasion is proven. Does this promise of confidentiality affect
compliance? There is very little empirical evidence of the impact of the confidentiality contract
on compliance in any one country and none across countries. Using identical laboratory
experiments conducted in the United States and Italy, we analyze the impact on tax compliance
of “Full Disclosure” (e.g., release of photos of tax evaders to all subjects, along with information
on the extent of their non-compliance) and of “Full Confidentiality” (e.g., no public
dissemination of photos or non-compliance). We find that compliance is greater both in the U.S.
and in Italy when there is public disclosure of information about individuals found to be tax
evaders.
JEL Classifications: H2, H3.
Keywords: Tax compliance, experimental economics, confidentiality, social norm.
* Tulane University, University of Venice Ca’Foscari, Georgia State University, Georgia State
University, and Georgia State University/African Tax Institute University of Pretoria. Please
address all correspondence to: Sally Wallace, Suite 600 Andrew Young School of Policy
Studies, Georgia State University, 14 Marietta Street NW, Atlanta, GA 30303-3992
(swallace@gsu.edu).

103.4.95.4 on Mon, 17 Jul 2023 05:02:30 +00:00
1. Introduction
Tax administrations are constantly looking for innovative ways to increase tax
compliance. Traditional compliance-inducing measures include penalty rates and audit
structures. A novel method that has been increasingly discussed is limited disclosure of taxpayer
information in cases of tax evasion. The threat of public shaming through disclosure adds an
additional non-financial penalty that may induce taxpayers to increase compliance to keep their
names clean. However, the threat of public disclosure could crowd out the intrinsic motivation
for compliance, causing a backlash against an intrusive government and reducing compliance as
a retaliatory action; additionally, with increasing sentiment against taxation, disclosure may
actually increase the utility for a subset of the population, an effect that may be reinforced if
“contagion” effects exist (possibly due to establishing a new social norm associated with reduced
compliance) such that observing that others have underreported income may reduce one’s own
compliance. Therefore, whether public disclosure of taxpayer compliance behavior increases or
decreases compliance is largely unknown. This paper uses a laboratory experiment to examine
the impact of confidentiality of taxpayer information on the level of individual compliance in
two countries in which baseline taxpayer compliance is arguably different – Italy and the United
States.
The impact of explicit disclosure of evasion has seldom been empirically studied at the
individual level due largely to the absence of reliable micro-level taxpayer data. Using field data,
Slemrod, Thoresen, and Bø (2012) utilized a natural experiment in Norway, in which after 2001
some tax data were made available on the internet while prior to 2001 such information was
available for only a select number of communities. They found on average a slight increase in
reported business income after 2002 in communities that previously had limited disclosure. Also,

103.4.95.4 on Mon, 17 Jul 2023 05:02:30 +00:00
Hasegawa et al. (2013) analyzed disclosure of individual and corporate tax information in Japan,
and found that the existence of a “disclosure threshold” encouraged some underreporting of
income.
Several laboratory experiments have looked at the effects of disclosure on taxpayer
compliance, with mixed results. Laury and Wallace (2005) conducted a laboratory experiment
that implemented a mild form of disclosure, in which the tax reports of a subset of participants
were displayed to other participants, as coded by an anonymous ID number, and they found
some suggestive evidence that disclosure has a positive effect on compliance. Fortin, Lacroix,
Villeval (2007) also studied the effects of feedback on tax reporting decisions. In their design,
subjects were told the number of subjects who underreported income in the previous round and
the mean level of reported income (but not the level of income reported by any individual). They
found that reported income was slightly lower when subjects received information on others’
reporting behavior, but also that an increase in the average level of evasion in the group was
associated with an increase in individual reported income. Lefebvre et al. (2011) compared tax
reporting behavior across three countries (France, Belgium, and the Netherlands), analyzing the
effects of providing subjects with a “good” example (e.g., the maximum proportion of subjects
who reported truthfully) or a “bad” example (e.g., the minimum proportion). They found that
subjects who observed “bad” examples of others’ behavior were less likely to fully report
income, but that reporting was largely unaffected by observing “good” examples. They also
found differences in reporting across countries, with underreporting more common in France and
the Netherlands than in Belgium. Also, Coricelli et al. (2010) focused on the emotional impact of
cheating and disclosure in a tax-reporting experiment. In a “pictures” treatment, a subject who
was audited and found to have unreported income had his or her photo shown to others in the

103.4.95.4 on Mon, 17 Jul 2023 05:02:30 +00:00
session. They find that the use of photos increased compliance (and also “emotional arousal”, as
measured by skin conductance responses); they also found higher compliance (and higher
emotional arousal) after an audit.
Despite these contributions, the impact of confidentiality versus disclosure of taxpayer
information on the level of individual compliance remains unknown. We seek to fill this gap by
using laboratory experiments to examine the impact of disclosure in two quite different
environments – the U.S. and Italy. In our experimental design, an individual is given income, and
then must decide how much of the income to report. Taxes are paid on reported income at a
preannounced tax rate, and no taxes are paid on unreported income. However, unreported income
may be discovered via an audit, and the subject must then pay the unpaid taxes plus a fine based
on the unpaid taxes. We introduce two main treatments into this system. In one treatment (“Full
Confidentiality”), an individual who is detected evading is financially penalized, but their
reporting information is not shared with the other subjects. In a second treatment (“Full
Disclosure”), those who have been caught evading find their non-compliance information shared
among the subjects via the display of their pircture on the computer screens of all subjects (along
with information on the level of underreporting).
While our Full Disclosure treatment mirrors the pictures treatment used by Coricelli et
al., our design differs in several important dimensions. These differences provide an additional
way in which disclosure may affect reporting decisions. Coricelli et al. intentionally
implemented a design that highlighted the role of social emotions (‘arousal’) and minimized
stigma associated with one’s photo being shown as a tax evader. In their experiment the
probablity of an audit is endogenous: the audit probability is highest for those who report the
least income; thus those who report less income decrease the likelihood that others (who report

103.4.95.4 on Mon, 17 Jul 2023 05:02:30 +00:00
more income) face a high audit probability.1 In contrast, the audit probabllity is exogenous in our
experiment and any subject with underreported income discovered via an audit has her picture
shown to everyone in the session. A participant’s reporting decision has no an external effect on
others’ audit probability. Thus, we can study whether (as policy-makers hope) the threat of social
stigma influences reporting decisions.
Importantly, we conduct separate but identical experiments in the United States and in
Italy, thereby providing us with quite different baselines of compliance norms. By performing
mirror experiments in these two countries, we believe that we are better able to identify the
marginal impact of evasion disclosure on compliance behavior than would be possible by
focusing on just one country.
We find strong support for the notion that public shame is an additional deterrent to tax
evaders, beyond the traditional enforcement tools of higher audit rates and enhanced penalty
rates. This effect that seems equally strong in the U.S. and in Italy, despite what appears to be a
different social norm of compliance in the two countries. We also find complicated interaction
effects between disclosure and other policy variables, effects that nonetheless confirm that
disclosure encourages compliance.
2. The Institutional Context
The United States and Italy present very different institutional perspectives on the role of
confidentiality in tax disclosure. This section discusses the different attitudes toward disclosure –
and toward tax compliance – in these two countries.
1
In addition, at most one subject’s photo is shown to others. If more than one participant is audited and found to
have under-reported income, the photo shown to others is randomly selected from those audited with under-reported
income.

103.4.95.4 on Mon, 17 Jul 2023 05:02:30 +00:00
In the United States, announcing the names of delinquent taxpayers is a departure from
the standard of practice for the federal government with respect to individual taxpayer data.
Confidentiality of individual taxpayer data is a long-held basic right of the U.S. system of tax
administration. Section 6103 of the Internal Revenue Code sets the guidelines for confidentiality
and for the limited disclosure of return information to state and local tax officials. As noted by
former Internal Revenue Service (IRS) Commissioner Margaret Richardson, “IRS employees are
prohibited from accessing information not needed to perform their official tax administration
duties” (Testimony, 15 April 1997). Confidentiality of taxpayer data is thereby guaranteed within
the system of tax administration, and the IRS imposes strict disclosure rules for individual
taxpayer data flowing outside the federal system to state tax administrators, other U.S.
government agencies, individuals, and companies. The IRS also imposes penalties for
unwarranted disclosure.
While taxpayers may believe that IRS rules ensure that tax information is largely private
and held in confidence by the IRS, the confidentiality of taxpayer data and information has not
always been a given. Until the mid-1970s, tax returns of publicly traded companies were
available to the public at-large. Some U.S. states have utilized a “wall of shame” approach to
increase voluntary compliance. For example, the revenue code of the state of Georgia allows
limited disclosure in certain cases of tax arrears. In West Virginia disclosure of corporate income
tax returns may occur once disputes in liabilities reach the point of the circuit court. Taxpayers
themselves have sometimes voluntarily chosen to make their returns public; indeed, many in
political office choose to do so.
The level of disclosure and taxpayer reaction therefore falls along some continuum from
subtle to extremely overt. At one end there is “administrative disclosure” in the form of sharing

103.4.95.4 on Mon, 17 Jul 2023 05:02:30 +00:00
information about taxpayers among units within government. At a more extreme end is
publication and announcement of names of individuals (and companies) who have tax arrears,
who fail to file tax returns, or who fail to make timely tax payments.2
In contrast, Italy is a country that offers a much more mixed attitude towards tax evasion
and tax evaders. On one side, Italy is often described as a country where the problem of tax
evasion is particularly acute.3 Estimates from various sources indicate that evasion in Italy is
much higher than in other highly developed countries (Giovannini, 2011), and evasion is often
considered at the root of many problems of the Italian economy: revenue losses, equity concerns,
and economic inefficiencies (Santoro, 2010). Italians are well aware of these problems. The
general sense of alarm was expressed well by former Prime Minister Mario Monti serving in a
government supported by a large political coalition, who remarked that “…against tax evasion,
Italy is in a state of war” (Mario Monti, Il Sole 24 ore, 17 August 2012). On the other side,
notwithstanding this widespread concern, the attitude of Italians towards tax evaders is more
tolerant than one might expect. For example, findings from social surveys consistently show that
Italians report an index of tax morale significantly lower than in many other countries (Alm and
Torgler, 2006).4 In a recent study of Italians’ opinions on tax evasion, Cannari and D’Alessio
(2007) find that aversion to tax evasion turns out to be quite low across all social classes, and for
this reason they argue that in Italy a mechanism of general reprobation may have only a modest
impact in reducing evasion.
2
Note that some have expressed concern that disclosure in its various forms represents a breach of the taxpayer-tax
administration confidentiality that is considered important in tax administration. However, Mazza (2003) argues that
disclosure and privacy concerns are not inextricably linked in such a way that all disclosure yields a breach of
privacy; see also Lenter, Shackelford, and Slemrod (2003).
3
As an example, see the recent article from The Economist Blog, which starts by noting that “Death may be certain
in Italy, but taxes are another matter: an estimated of € 285 billion remained unpaid last year, about 18% of GDP”
(http://www.economist.com/blogs/schumpeter/2013/01/tax-evasion-italy ).
4
As discussed later, it is of course possible that evasion is high precisely because tax morale is low.

103.4.95.4 on Mon, 17 Jul 2023 05:02:30 +00:00
The ambivalence Italian attitudes towards tax evasion also emerges from the way in
which the issue of disclosure of taxpayers’ information is treated by the law and perceived by the
public. In principle, the Italian law allows for ample disclosure of individual taxpayer data. In
particular, the law establishes that every year the tax administration compile lists with the names
of all Italian taxpayers, their total income, taxable income, and major source of income. The lists
are then made available by the Italian Tax Agency (“Agenzia delle Entrate”) and the taxpayers’
municipalities to anyone who is interested. The standard practice has for many years been that
local and national newspapers occasionally access the lists and publish the names and the
incomes of wealthier people. This sometimes occurs for people found to be tax evaders as well.
An issue that has raised much concern recently is whether the Agency or the municipalities can
by themselves publish this information. The problem exploded in 2008 when the Tax Agency
published an internet list of all Italian taxpayers for everyone to access directly from his or her
computer. This decision by the Agency inflamed public opinion, which was split between
privacy advocates and supporters who argued that the publication of the data could finally
undermine the prevailing system of tax evasion. The list remained available only for few hours
since the Italian Data Protection Authority declared the publication illegitimate, complaining
both of the lack of explicit permission in the Italian law and of a more general problem of the
disparity between the need for public transparency on taxpayer data and the ability to make them
available in the web.
Following this episode, there have since been other discussions in the media and in
politics on the opportunity to make individual taxpayers’ data more generally available, but

103.4.95.4 on Mon, 17 Jul 2023 05:02:30 +00:00
without much effect.5 As a result, the situation in Italy is that information regarding taxpayers
and tax evaders is only occasionally made available by local and national newspapers.
3. Theoretical Background
Following Allingham and Sandmo (1972), tax compliance is typically modeled as a
decision of how much income to report to tax authorities given that underreporting may be
discovered with some audit probability and penalized with some penalty rate.6
To illustrate this approach more precisely, consider a simple version of the standard
model. An individual is assumed to receive a fixed amount of income I, and must choose how
much of this income to declare to the tax authorities and how much to underreport. The indi-
vidual pays taxes at rate t on every dollar D of income that is declared, while no taxes are paid on
underreported income. However, the individual may be audited with a fixed probability p; if
audited, then all underreported income is discovered, and the individual must pay a penalty at
rate f on each dollar that he was supposed to pay in taxes but did not pay. The individual's
income IC if caught underreporting equals
(1) IC=I-tD-f[t(I-D)],
or income less taxes paid on reported income less penalties on unreported taxes. If
underreporting is not caught, income IN is
(2) IN=I-tD,
5
Some recent laws establish that politicians and public managers have to make publicly available their fiscal returns
(L. 190/2012 and Dlgs. 33/2012). The laws have however been introduced more as an anti-corruption policy than as
way to fight tax evasion. Moreover, it is not clear what sanctions actually apply for those who do not comply and,
indeed, the applications of the dispositions have been so far carried out mainly as a voluntary decision.
6
See Cowell (1990), Andreoni, Erard, and Feinstein (1998), Alm (1999), and Slemrod and Yitzhaki (2002) for
comprehensive surveys of the evasion literature. See Alm (2012) and Sandmo (2012) for more recent discussions.

103.4.95.4 on Mon, 17 Jul 2023 05:02:30 +00:00
or income less taxes paid on reported income. The individual is assumed to choose declared
income to maximize expected utility, defined as
(3) EU(I)=pU(IC )+(1-p)U(IN ),
where E is the expectation operator and utility U(I) is a function only of income. This
optimization generates a standard first-order condition for an interior solution; given concavity of
the utility function, the second-order condition is satisfied. Comparative statics results are easily
derived. For example, it is straightforward to show that an increase in the probability of detection
p and the penalty rate f unambiguously increase declared income.
This economics-of-crime approach therefore gives the sensible result that compliance
depends upon enforcement. This approach also concludes that an individual pays taxes because –
and only because – of the economic consequences of detection and punishment. This is a
plausible insight, with the obvious implication that the government can encourage greater tax
compliance by increasing the audit and the penalty rates.
However, it is clear to many observers that compliance cannot be explained entirely by
such purely financial considerations, especially those generated by the level of enforcement
(Graetz and Wilde, 1985; Elffers, 1991; Kirchler, 2007; Slemrod, 2007; Torgler, 2007). The
percentage of individual income tax returns that are subject to a thorough tax audit is generally
quite small in most countries, almost always well less than 1 percent of all returns. Similarly, the
penalty on fraudulent evasion seldom exceeds more than the amount of unpaid taxes, and these
penalties are infrequently imposed; civil penalties on non-fraudulent evasion are even smaller. A
purely economic analysis of the evasion gamble suggests that most rational individuals should
either underreport income not subject to source withholding or over-claim deductions not subject
to independent verification because it is unlikely that such cheating will be caught and penalized.
10

103.4.95.4 on Mon, 17 Jul 2023 05:02:30 +00:00
Yet even in the least compliant countries evasion seldom rises to levels predicted by a purely
economic analysis, and in fact there are often substantial numbers of individuals in most
countries who apparently pay all (or most) of their taxes all (or most) of the time, regardless of
the financial incentives they face from the enforcement regime.
The puzzle of tax compliance behavior may therefore be why people pay taxes, not why
they evade them. This observation suggests that the compliance decision must be affected in
ways not captured by the basic economics-of-crime approach. Motivations affected by
confidentiality may well play a role here.
Indeed, it is possible to extend the basic economics-of-crime approach to incorporate the
potential role of confidentiality in the individual compliance decision. Perhaps the simplest
extension is to introduce the role of “ethics”. There is much evidence of what may be termed a
“social norm” of tax compliance. Although difficult to define precisely, a social norm can be
distinguished by the feature that it is process-oriented, unlike the outcome-orientation of
individual rationality (Elster, 1989). A social norm therefore represents a pattern of behavior that
is judged in a similar way by others and that therefore is sustained in part by social approval or
disapproval. Consequently, if others behave according to some socially accepted mode of
behavior, then the individual will behave appropriately; if others do not so behave, then the
individual will respond in kind (Frey and Torgler, 2007). The presence of a social norm is also
consistent with a range of approaches, including those that rely upon social customs, even
appeals to patriotism, conscience, or upon an individual’s feelings of altruism, fairness, morality,
guilt, or alienation.
Overall, this factor of taxpayer ethics, broadly defined, suggests that an individual will
comply as long as she believes that compliance is the “right thing to do”. Conversely, if
11

103.4.95.4 on Mon, 17 Jul 2023 05:02:30 +00:00
noncompliance becomes pervasive, then the ethics of compliance disappears. This perspective
also suggests that disclosure of one’s compliance choices to others may affect the social norm of
compliance and, through this channel, each individual’s compliance decision.
There are several ways in which the role of ethics can be introduced in the model of self-
interested individual behavior. Perhaps the most straightforward way is suggested by Kahneman
and Tversky (1979), who incorporate what they term a “reference point” as a form of social norm in
prospect theory. They assume that a loss in utility occurs if individuals do not achieve some
reference point, a phenomenon they call “loss aversion”. The loss may be avoided by reporting all
income and paying all taxes; individuals who declare less than their full income and pay less than
their full taxes will suffer a loss in utility.
More formally, assume that each individual maximizes expected utility, where income in the
two states of the world is now defined not as in equations (1) and (2) but as
(1)’ IC=I-tD-f[t(I-D)]-γt(I-D)
(2)’ IN=I-tD-γt(I-D).
Expected utility is still defined by equation (3). The individual now is assumed to suffer a
psychological loss in expected income proportional to undisclosed taxes, where the coefficient γ
measures as a fraction how much the individual would pay to avoid the loss associated with each
dollar of unreported taxes. It is straightforward to show that declared income is higher in this setting
than in the basic economics-of-crime model discussed earlier.
Clearly, γ is likely to be sensitive to the disclosure of one’s compliance behavior to others;
that is, γ is likely to vary with the amount of confidentiality. This effect can work through several
channels: social stigma, personal and cultural values (including religion and/or ideology), the
influence of peers, the perceived quality of fiscal institutions, and the equity of the fiscal system.
12

103.4.95.4 on Mon, 17 Jul 2023 05:02:30 +00:00
Through social stigma, γ is likely to be sensitive to the public disclosure to others of one’s
compliance behavior; that is, γ is likely to vary with the amount of confidentiality. The stronger the
ethical norm to pay one’s taxes fully, the more deviant the behavior of a non-compliant individual
becomes, and the more loss the individual feels with disclosure. Cultural and personal values can
also affect the strength of the norm under disclosure. For example, the effect of public disclosure
could be weaker in societies where private values like family and friendship are more important,
while disclosure could be stronger in societies in which civic values like justice and politics are
rated higher. Religions that encourage forgiveness may reduce the effect of disclosure on
compliance, while faiths that put less weight on mercy and more on individual responsibility could
enhance the effect of disclosure.
Of course, disclosure could instead lead to a decline in γ if noncompliance is viewed as a
legitimate form of government protest. There are several reasons why this may occur. Some work
has emphasized that, if taxpayers perceive that services provided by the public sector in return for
taxes are not “fair”, then taxpayers may respond by increasing evasion as a form of civic protest
(Mason and Calvin, 1984; Cowell and Gordon, 1988; Bordignon, 1993; Rablen, 2010). In this case,
public disclosure could make taxpayers even more aggressive, which in the context of equations
(1)’ and (2)’ implies a lower value for γ. The intuition is straightforward. If taxpayers do not
believe that the government services provided are commensurate with tax payments, then public
disclosure increases their mistrust/dislike for government and they respond by reducing
compliance as a statement against government.
A decline in γ following disclosure could also be due to the interaction with ethics. The
impact of taxpayer ethics suggests that an individual will comply as long as he or she believes
that compliance is the “right” thing to do; conversely, if noncompliance becomes pervasive, then
13

103.4.95.4 on Mon, 17 Jul 2023 05:02:30 +00:00
the ethics of compliance may disappear (Myles and Naylor, 1996; Fortin, Lacroix, and Villeval,
2007; Traxler, 2010). This logic implies that γ decreases with the evasion of others. It also
implies that disclosure of one’s compliance choices to others may affect the social norm of
compliance and, through this channel, each individual’s compliance decision. In other words, the
social norm may be affected via a “contagion” effect such that observing that others have
underreported income may reduce the overall compliance rate.
This discussion illustrates the complexity of the compliance effects of breaking
confidentiality. As emphasized earlier, the actual evidence on confidentiality versus disclosure is
not clear-cut. The next section presents our experimental design to examine this issue.
4. Experimental Design
We use a laboratory experiment to analyze the impact of announcing tax evasion on the
overall level of tax compliance in Italy and in the United States. Experimental methods have long
been used to study compliance. They allow many factors suggested by theory to be introduced in
experimental settings, they allow these factors to be introduced singly and exogenously in a
controlled environment, and they generate precise data on individual compliance behavior. There
are also legitimate concerns regarding small sample sizes and the use of student subjects,
concerns that relate to the “external validity” validity of laboratory experiments. Still, there is
now much evidence that there is little difference between student and nonstudent responses in most
environments (Plott, 1987), including evidence that relates directly to compliance behavior (Alm,
Bloomquist, and McKee, 2015). Most importantly, there is now a large literature (Smith, 1976,
1982) that argues convincingly that experimental methods can contribute significantly to policy
debates, as long as some conditions are met: the payoffs to subjects must be salient, better subject
14

103.4.95.4 on Mon, 17 Jul 2023 05:02:30 +00:00
decisions yield higher subject payoffs, decision costs must be commensurate with the payoffs, and
the experimental setting must capture the essential properties of the naturally occurring environment
that is the subject of investigation. These conditions are met in our experimental design.
Our basic experimental design follows that typically used in previous tax compliance
experiments (Alm and Jacobson, 2007, Alm, 2010, Laury and Wallace, 2005).7 An individual is
given income, and then must decide how much of the income to report. Taxes are paid on
reported income at a preannounced tax rate of 30 percent, and no taxes are paid on unreported
income. Unreported income may be discovered via an audit. If the subject is audited and found to
have underreported income, the subject must then pay the unpaid taxes plus a fine based on the
unpaid taxes at a preannounced fine rate.
Figure 1 shows a sample screenshot from our experiment. The computer interface
displays the tax rate, probabiility of an audit, penalty rate, income, and tax owed (income
multiplied by the tax rate).8 Subjects used the scroll bar to enter a tax reporting decision, with the
chosen amount of income (and associated tax payment) shown in the boxes above the scroll bar.
The subject’s earnings (whether or not audited) were also shown for any level of income
indicated. Subjects were able to revise their choice by sliding the scroll bar to different levels of
income. The resulting earnings (if audited or if not audited) were updated and displayed for any
level of income selected. Thus the earnings implications for any possible reporting decision were
transparent and easily available to the subject. The choice was only finalized when the “File
Taxes” button was clicked. After all subjects did so, the audit outcome was randomly determined
independently for each individual according to the pre-announced audit probability. The
7
A copy of the experimental instructions is found in Appendix A.
8
The countdown timer was displayed on subject’s computer screen but the time limit was not enforced. Most
reporting decisions were made quickly (almost always before the timer reached zero). In those cases when the
expired subjects still were able to enter their decision. The tax history button allowed a subject to view her own tax
reporting decisions, audit outcome, and period earnings from all prior rounds of the experiment.
15

103.4.95.4 on Mon, 17 Jul 2023 05:02:30 +00:00
computer screen then displyed the subject’s earnings for that period (see the top panel of Figure
2).
Each subject chose how much income to report in 16 rounds. Period earnings were
revealed at the end of each round. However, the subject’s final earnings for the experiment were
based soley on the earings from one round. The round used for payment was randomly selected
at the end of the session based on a single throw of a 16-sided die. Subjects were then privately
paid their earnings from this one round only (plus a fixed participation payment). All decisions
were conducted on computers, with each subject assigned to his or her own computer. All
computer screens were visually isolated from the others’ computers.
Income was held constant ($25 in the U.S. and €15 in Italy), as was the tax rate (30
percent). The probability of an audit was either 20 percent or 30 percent. The fine on unreported
income was either 100 percent (so that the individual paid unpaid taxes plus an additional
penalty equal to the amount of unpaid taxes) or 200 percent (or the individual paid unpaid taxes
plus a penalty equal to twice the amount of unpaid taxes). Thus there were four different
treatment combinations based on the audit probability and penalty. Each treatment combination
was held fixed for four rounds before a new treatment combination was presented to subjects.
The order of treatments was reversed between sessions. In addition to their earnings from one
randomly-selected round, subjects received a fixed participation payment ($10 in the U.S. and €7
in Italy). Experiments lasted about 90 minutes, and earnings ranged from $30 to $55. 9
The main policy variable is the impact of public disclosure versus confidentiality on tax
compliance. We consider two very different levels of confidentiality. In a “Full Confidentiality”
treatment, all information is private: subjects who are audited and found to be noncompliant are
9
We set the period-income and participation payment levels so experiment earnings relative to average local
student-employment wages were roughly equal across the two countries.
16

103.4.95.4 on Mon, 17 Jul 2023 05:02:30 +00:00
assessed a penalty; however, none of their compliance information is shared among the other
subjects. In a second treatment (“Full Disclosure”), subjects who are audited and caught evading
are assessed the financial penalty, but their tax-reporting information is no longer kept private.
At the end of each round, all participants were shown photos of anyone who was audited and had
underreported income. The amount of income reported by each subject was shown with the
subject’s photo. A sample screenshot is shown in the bottom of Figure 2.10
The experiments were conducted at Georgia State University in the United States and at
the University of Venice Ca’Foscari in Italy using student subjects from each institution. The
subjects also filled out a demographic survey at the end of the experiment, which helps us to
capture socioeconomic and demographic data that are used in the empirical analysis.
5. Analysis and Results
In total, there were 2,720 individual observations (16 decisions made by each of 170
subjects); 92 subjects participated in the US (1,472 individual observations) and 78 in Italy
(1,248 individual observations). Table 1 summarizes the descriptive statistics for the subjects in
both countries. The average age of the subjects was 23.8 years (22.5 years, U.S.; 25.1 years,
Italy), and 43.5 percent of the subjects were male (39.1 percent, U.S.; 48.7 percent, Italy). The
racial and religious mix of the subjects varied significantly between the U.S. and Italy. In the
U.S. sessions, 75.0 percent of the subjects were black and 6.5 percent were Catholic; in the Italy
sessions, 2.5 percent were black and 62.8 percent were Catholic.
Turning to our primary variable of interest, there are several ways in which we can
measure the level of compliance of our subjects. One measure is a binary variable that represents
10
The photos shown in Figure 2 are not those of any subjects in our experiment. All photos were permanently
deleted from the camera and computer files immediately after each session.
17

103.4.95.4 on Mon, 17 Jul 2023 05:02:30 +00:00
whether the subject reported all income or not (equal to 1 if the subject reported all income and 0
otherwise). Any underreported income, no matter its size, is considered non-compliance. Across
all treatments and experiment parameters, subjects fully reported income 48.2 percent of the
time. An alternative measure defines compliance as the proportion of income that is reported, or
the average Compliance Rate. On average, subjects reported 71.3 percent of their income (See
Table 2). Unless otherwise stated, we use this average compliance rate as our measure of
compliance, although our results are largely robust to either measure.
Table 3 presents summary information on the average compliance rate by treatment and
by country. Pooling across sessions, subjects respond in a predictable manner to our experiment
parameters (probability of an audit and penalty rate). In both the U.S. and Italy, mean reported
income generally increases as the probability of an audit increases and as the penalty on
unreported income increases. The only exception to this is in treatments where the mean level of
reporting is highest (or over 70 percent of income reported in the U.S., and over 80 percent of
income reported in Italy). In these treatments, changes in the experiment parameters had little
effect on the amount of income reported.
These simple descriptive statistics in Table 3 allow us to test several hypotheses. Our
main interest is in the effect of changes in the level of confidentiality on compliance and on the
extent to which reporting behavior differs across countries. We test several hypotheses that
explore how confidentiality, culture and related variables affect compliance. We address each in
turn.
Of most interest is the effect of Full Confidentiality versus Full Disclosure, and a first
hypothesis examines the impact of disclosure, in which photos and underreported income are
18

103.4.95.4 on Mon, 17 Jul 2023 05:02:30 +00:00
shown for those subjects found to be noncompliant in an audit, where the variable Photos equals
1 in sessions where pictures are shown of noncompliant subjects and 0 otherwise:
Hypothesis 1: Showing photos of subjects who are audited and have underreported income will
increase tax compliance.
Overall, our data support this hypothesis. Pooling data across countries (and across experiment
parameters), 68.9 percent of income is reported in baseline sessions, and 73.1 percent in sessions
where photos are shown of those with underreported income. If we use our binary measure of
compliance, 37.7 percent fully report in the baseline condition compared with 56.5 percent in the
sessions with photos. These differences are significant at any standard level of confidence and
provide support for Hypothesis 1.
A second hypothesis examines the impact of public disclosure in the U.S. versus Italy:
Hypothesis 2: The photo treatment will have a larger effect on compliance in the U.S. than in
Italy.
If underreporting income is more acceptable in Italy than in the U.S., then one might expect that
lowering the level of confidentiality will have a larger effect in the U.S. than in Italy. Our data
support this hypothesis. As seen in Table 3, compliance is significantly higher in Italy than in the
U.S. in both the baseline and photos treatments. However, the marginal impact of the photo
treatment causes only a modest increase in compliance in Italy (from 73.1 to 74.7 percent)
compared to the U.S. (from 64.8 to 71.9 percent).
Another hypothesis pertains to the impact of seeing other individuals who are non-
compliant (a contagion effect of sorts):
Hypothesis 3: Observing more subjects underreporting income will lead to lower compliance
rates.
19

103.4.95.4 on Mon, 17 Jul 2023 05:02:30 +00:00
We restrict our attention to the photo treatment, and consider the effect of the photos treatment
on individual-level compliance (holding other effects constant). We address this hypothesis by
estimating the following model:
CRi,t = 1+2Treatmentt+3Low Auditt+4Low Penaltyt+5Italyt+6Xi+ψt+ui+εi,t (4)



where the dependent variable CRi,t denotes subject i’s Compliance Rate in period t; Treatmentt is
a dummy variable for periods in which there is some policy innovation (e.g., public disclosure
via photos, compliance behavior of non-compliant subjects); Low Auditt is a dummy variable for
the audit rate, equal to 1 if the audit probability is high and 0 otherwise; Low Penaltyt is a
dummy variable for the penalty rate, equal to 1 if the penalty rate is high and 0 otherwise; Italyt
is a dummy variable equal to 1 for sessions run in Italy and 0 otherwise; Xi is a vector of
demographic variables (e.g., subject sex and religion); ψt is a set of T-1 dummies that capture
potential non-linear period effects; ui are random effects that control for unobservable individual
characteristics; εi,t is the contemporaneous additive error term; and k is the coefficient for
variable k. The model is estimate using Tobit maximum likelihood methods with random
effects.
As shown in Table 4, we do not find support for this hypothesis. While being audited and
found to have underreported income in the previous round has a significant negative effect on
compliance, an increase in the proportion of others in a session that were shown to have
underreported income has no significant effect on compliance. Most other variables are
significant. For example, note that compliance is significantly higher in Italy than in the U.S. and
lower for male subjects.
20

103.4.95.4 on Mon, 17 Jul 2023 05:02:30 +00:00
Several additional hypotheses pertain to the impact of enforcement parameters on
compliance:
Hypothesis 4: An increase in the audit rate will increase compliance.
Hypothesis 5: An increase in the penalty rate will increase compliance.
These hypotheses are largely supported by the estimation results in Table 4. For example,
compliance is lower with a lower audit probability and a lower penalty for underreported
income.
Another hypothesis relates to the interaction effects of disclosure and audit rates:
Hypothesis 6: The effect of photos may be stronger in sessions with a low probability of being
audited.
If compliance is subject to crowding-out, there may be less room for reducing confidentiality to
increase compliance because baseline compliance rates are already high in sessions with a high
audit probability (or a high penalty on underreported income). The simple descriptive statistics in
Table 3 lend some support to this hypothesis. The photo treatment has a smaller effect when the
audit probability and penalties are high.
We look at this more formally in Table 5, where we modify the basic empirical model to
include both penalty-rate controls and audit-probability interacted with photos (equal to 1 in
sessions with a photo), again estimated with Tobit maximum likelihood methods with random
effects. The baseline estimations in models (1) and (2) confirm that compliance is higher with
photos, and lower under low audit probabilities. As we move across the table to examine
crowding-out effects, it appears that the effects of the photos depend on the probability of audit
21

103.4.95.4 on Mon, 17 Jul 2023 05:02:30 +00:00
as well. In particular, pictures are about half as effective in a high audit environment. This result
is robust to demographic controls.
Although showing photos of those who have been audited and discovered to have
underreported income increases overall compliance, once a person has been identified to others
as a tax evader the stigma of further identification may be diminished. If this is the case, then one
would expect less compliance from those who were previously identified. Table 6 explores this
hypothesis:
Hypothesis 7: The effect of photos on compliance is smaller after one has been identified to
others as having been non-compliant.
We examine this issue both with Tobit estimation and with an alternative estimation method, a
double hurdle model. Tobit estimation assumes that the determinants of tax compliance have the
same effect on whether a person evades and (conditional on evasion) on how compliant the
person is; that is, people who choose not to evade are fundamentally the same as those who do.
However, if the social costs of evasion have a fixed component, this may be an unreasonable
assumption. As such, we follow Cragg (1971) and implement a double-hurdle model to relax the
assumption of equivalent effects on the intensive and extensive margins, estimated with probit
methods. These estimation results are presented in Table 6. Note that the dependent variable is
redefined as [1-Compliance Rate].
In the first model, we simply relax the Tobit restrictions, and examine the effect of a
photo treatment on compliance. In this model any stigma cost of pictures comes exclusively on
the extensive margin. As we move horizontally across Table 6, we add complexity to each
specification, by adding dummy variables to control for treatment parameters such as the
probability of audit, the size of the penalty, the country in which the sessions were run, and
22

103.4.95.4 on Mon, 17 Jul 2023 05:02:30 +00:00
various interaction terms. Regardless, in each specification the result remains on the extensive
margin; that is, subjects in a photo treatment are as much as 130 percent less likely to evade at
all. However, conditional on evading, being in a photo treatment has no negative deterrent effect.
In fact, conditional on evasion, those in photo treatments are less compliant than their baseline
counterparts.
While this result seems self-explanatory, there are some additional factors we must
consider. First, following the logic that there exists a fixed stigma cost, one interesting question
is what do people do given they have been previously caught evading. In Table 7 we add
complexity to the hurdle model (Table 6). Specifically, we re-run the model conditional on how
many times the subject has ever been caught, with times caught ranging from ever (>0) to at
least 4 (>3). Given that a subject has been caught at least once, we see a comparable, albeit
smaller, effect on tax evasion as above. However, this effect becomes insignificant as a subject is
caught additional times and, though still insignificant, we eventually see the sign of the effect
reverse (see Table 7).
Additionally, these hurdle models have used a pooled cross-section data structure.
However, given the 16 consecutive rounds of the experiment, our data actually exist as a panel.
We explore the implications of this design in Table 8. Here we replicate hurdle models 1-4 from
Table 6 with the difference that these new models are run in the context of a random-effects
panel. Again, we find that the effect of photos comes through in the first hurdle, with a similar
magnitude; that is, subjects are about twice as likely to not evade, in photos treatments and in the
panel model, and this effect only comes through in the first hurdle. Further, in each specification,
the error terms between the first and second hurdles are uncorrelated since the transformed Rho
is never significant, suggesting there is no need to use a bootstrap technique.
23

103.4.95.4 on Mon, 17 Jul 2023 05:02:30 +00:00
Finally, we ask why we only observe this effect on the extensive margin. In principle,
there are two distinct possibilities. First, conditional on evasion, the evasion rates in the photo
treatments are characterized by a right-shift of probability mass with respect to the baseline
treatment at all levels of evasion. Second and otherwise, conditional on evasion, the evasion rates
in the photo treatments are simply the left-truncated evasion rates in the baseline treatment. Note
that there are different policy implications of the two possibilities. In particular, under the latter,
while a policy of disclosure increases the number of those who do not evade at all, it also
increases the distance in after-tax incomes between those who evade and are not caught and
those who do not evade. The result therefore may increase horizontal inequity due to tax evasion.
Figure 3 presents the cumulative distribution function (CDF) of evasion rates by
treatment, conditional on evasion. It appears from Figure 3 that there is indeed a treatment effect
of showing pictures in both a shift of probability mass at all levels of evasion and a greater mass
of people evading fully. We confirm this hypothesis with a two-sample Kolmogorov-Smirnov
test for equality of distributions, which suggests smaller levels of evasion in the baseline
treatment (D=0.17, p=0.00).
6. Conclusions
How will individuals react to the public disclosure of their tax evasion? Will they treat
public shame as an additional cost of cheating and respond by increasing their compliance? Or
will the threat of public shame crowd out an individual’s intrinsic motivation to obey the law,
leading to a negative reaction against an intrusive government and reducing compliance as a
retaliatory action.
This paper uses laboratory methods to examine the impact of disclosure on tax
compliance, comparing subject responses in Italy and the United States. Our results provide
24

103.4.95.4 on Mon, 17 Jul 2023 05:02:30 +00:00
strong support for the notion that public shame is an additional deterrent to tax evaders, beyond
the traditional enforcement tools of higher audit rates and enhanced penalty rates. Relatively low
cost disclosure could yield an important increase in compliance, on the order of around 5 percent
in each country. This deterrence is equally strong in the U.S. and in Italy, despite what appears to
be a different social norm of compliance in the two countries. While deterrence via audits and
fines must remain as basic elements of any sustained government policy to improve compliance,
our results suggest that public disclosure of tax evaders represents an additional – and powerful –
instrument.
References
Allingham, M. and A. Sandmo (1972). Income tax evasion: A theoretical analysis. Journal of
Public Economics 1 (3-4): 323-38.
Alm, J. (1999). Tax compliance and administration. In W. B. Hildreth and J. A. Richardson

(eds.), Handbook on Taxation. New York, NY: Marcel Dekker, Inc., 741-768.
Alm, J. (2010). Testing behavioral public economics theories in the laboratory. National Tax
Journal 63 (4, Part 1): 635-658.
Alm, J. (2012). Measuring, explaining, and controlling tax evasion: Lessons from theory, field
studies, and experiments. International Tax and Public Finance 19 (1): 54-77.
Alm, J., K. M. Bloomquist, and M. McKee (2015). On the external validity of laboratory
experiments. Economic Inquiry.
Alm, J. and S. Jacobson (2007). Using laboratory experiments in public economics. National Tax
Journal 60 (1): 129-152.
Alm, J. and B. Torgler (2006). Culture differences and tax morale in the United States and in
Europe. Journal of Economic Psychology 27: 224-246.
Alm, James, G. H. McClelland, and W. D. Schulze (1992). Why do people pay taxes? Journal of
Public Economics 48 (1): 21-38.
Andreoni, J., B. Erard, and J. S. Feinstein (1998). Tax evasion. The Journal of Economic
Literature 36 (2): 818-860.
25

103.4.95.4 on Mon, 17 Jul 2023 05:02:30 +00:00
Blaufus, K., J. Bob, and P. Otto (2014). The Effect of Tax Privacy on Tax Compliance – An
Experimental Investigation, arqus Discussion Paper No. 164.
Cannari, L., and G. D’Alessio (2007). Le opinioni degli italiani sull’evasione fiscale. Banca
d’Italia, Temi di discussione, n. 618.
Coricelli, G., M. Joffily, C. Montmarquette, and M. Villeval (2010). Cheating, emotions, and
rationality: an experiment on tax evasion. Experimental Economics 13: 226-247.
Cowell, F. (1990). Cheating the Government: The Economics of Evasion. Cambridge, MA: The
MIT Press.
Cragg, John G. "Some statistical models for limited dependent variables with application to the
demand for durable goods." Econometrica: Journal of the Econometric Society (1971):
829-844.
Cummings, R. G., J. Martinez-Vazquez, M. McKee, M., and B. Torgler (2009). Tax morale
affects tax compliance: Evidence from surveys and an artefactual field experiment.
Journal of Economic Behavior and Organization 70 (3): 447-457.
Elffers, H. (1991). Income Tax Evasion: Theory and Measurement. Amsterdam, The
Netherlands: Kluwer.
Elster, J. (1989). The Cement of Society – A Study of Social Order. Cambridge, UK: Cambridge
University Press.
Fortin, B., G. Lacroix, and M. Villeval (2007). Tax Evasion and Social Interactions. Journal of
Public Economics 91(11-12): 2089-2112.
Frey, B. S. and B. Torgler (2007). Tax morale and conditional cooperation. Journal of
Comparative Economics 35 (1): 136-159.
Giovannini, E. (2011). Economia non osservata e flussi finanziari. Rapporto finale. Ministero
dell’Economia e delle finanze, Roma.
Graetz, M. J. and L. L. Wilde (1985). The economics of tax compliance: Fact and fantasy.
National Tax Journal 38 (3): 355-363.
Hasegawa, M., J. Hoopes, R. Ishida, and J. Slemrod (2013). The effect of public disclosure on
reported taxable income: Evidence from individuals and corporations in Japan. National
Tax Journal 66(3): 571-608.
Kahneman, D. and A.Tversky (1979). Prospect theory: An analysis of decision under risk.
Econometrica 47 (2): 263-292.
26

103.4.95.4 on Mon, 17 Jul 2023 05:02:30 +00:00
Kirchler, E. (2007). The Economic Psychology of Tax Behavior. Cambridge, UK: Cambridge
University Press.
Laury, S. and S. Wallace (2005). Confidentiality and taxpayer compliance. National Tax Journal
58 (3): 427-438.
Lefebvre, M., P. Pestieau, A. Riedl, and M. Villeval (2011). Tax evasion, welfare fraud, and “the
broken window” effect. Netspar Discussion Paper, DP02, 2011-110.
Lenter, D., D. Shackelford, and J. Slemrod (2003). Public disclosure of corporate tax return
information: Accounting, economics and legal issues. National Tax Journal 56 (4): 803–
830.
Mazza, S. W. (2003). Taxpayer privacy and tax compliance. University of Kansas Law Review
51: 1065-1144.
Plott, C. (1987). Plott, C. R. (1987). Dimensions of parallelism: Some policy applications of

experimental methods.” In A. E. Roth (ed.), Laboratory Experimentation in Economics:
Six Points of View. New York, NY: Cambridge University Press, 193-229.
Sandmo, A. (2012). An evasive topic: Theorizing about the hidden economy. International Tax
and Public Finance, 19 (1), 5-24.
Santoro, A. (2010). L’Evasione Fiscale. Bologna: Il Mulino.
Slemrod, J. (2007). Cheating ourselves: The economics of tax evasion. The Journal of Economic
Perspectives 21 (1): 25-48.
Slemrod, J., T. O. Thoresen, and E. E. Bø (2012). Taxes on the internet: Deterrence effects of
public disclosure in Norway. University of Michigan Working Paper. Ann Arbor, MI.
Slemrod, J. and S. Yitzhaki (2002). Tax avoidance, evasion, and administration. In A. J.

Auerbach and M. Feldstein (eds.), Handbook of Public Economics, Volume 3.
Amsterdam, London, and New York: Elsevier, 1423-1470.
Smith, V. (1976). Experimental economics: Induced value theory. The American Economic
Review Papers and Proceedings 66 (2): 274-279.
Smith, V. (1982). Microeconomic systems as an experimental science. The American Economic

Review 72 (5): 923-955.
Torgler, B. (2007). Tax Compliance and Tax Morale: A Theoretical and Empirical Analysis.
Cheltenham, UK: Edward Elgar.
27

103.4.95.4 on Mon, 17 Jul 2023 05:02:30 +00:00
Table 1: Descriptive Statistics
Variable US Italy
Observations 1472 1248
Age 22.5 25.1
Percent Black 75.0 2.5
Percent Catholic 6.5 62.8
Percent Male 39.1 48.7
Table 2: Descriptive Statistics of Decisions

Treatment Compliance Rate Standard Deviation Observations
All Sessions 71.3% 38.5 2720
Full Confidentiality 69.0 37.4 1200
Full Disclosure 73.1 39.2 1520
U.S. 69.0 40.6 1472
Italy 73.9 35.6 1248
Low Audit 66.9 39.9 1360
High Audit 75.6 36.5 1360
Low Penalty 66.4 40.2 1360
High Penalty 76.1 36.1 1360
Table 3: Average Compliance Rate by Treatment

Panel A: Italy Panel B: U.S.
Penalty: 100% Baseline Photos Penalty: 100% Baseline Photos
Audit Probability: 20% 56.1 62.2 Audit Probability: 20% 58.0 62.6
Penalty: 200% Baseline Photos Penalty: 200% Baseline Photos
28

103.4.95.4 on Mon, 17 Jul 2023 05:02:30 +00:00
Table 4: Tobit Estimation Results – Disclosure Effects
Model
Explanatory Variables (1) (2) (3)
Identified in Last Period with Photo -0.571*** -0.563*** -0.526***
(0.111) (0.110) (0.108)
Percent of Others Identified in Last Period 0.107 0.170 0.230
(0.588) (0.583) (0.571)
Low Audit -0.189*** -0.187***
(0.0688) (0.0673)
Low Penalty -0.227*** -0.223***
(0.0692) (0.0676)
Italy 0.180**
(0.0834)
Male -0.389***
(0.0685)
Catholic -0.279***
(0.0890)
Constant 1.234*** 1.438*** 1.595***
(0.0520) (0.0728) (0.0828)
Sigma 1.124*** 1.111*** 1.081***
(0.0481) (0.0474) (0.0461)
Observations 1,425 1,425 1,425
Standard errors are in parentheses; *** p<0.01, ** p<0.05, * p<0.1
29

103.4.95.4 on Mon, 17 Jul 2023 05:02:30 +00:00
Table 5: Tobit Estimation Results – Interaction Effects
Model
Explanatory Variables (1) (2) (3) (4)
Identified in Last Period with Photo 0.214*** 0.212*** 0.123** 0.181***
(0.0383) (0.0380) (0.0540) (0.0640)
Low Audit Probability -0.218*** -0.310*** -0.313***
(0.0377) (0.0550) (0.0544)
Photos * Low Audit 0.174** 0.176**
(0.0751) (0.0742)
Italy 0.123**
(0.0542)
Italy * Pictures -0.124*
(0.0746)
Low Penalty -0.233***
(0.0372)
Constant 0.870*** 0.980*** 1.027*** 1.081***
(0.0281) (0.0343) (0.0400) (0.0522)
Sigma 0.880*** 0.872*** 0.871*** 0.859***
(0.0233) (0.0230) (0.0230) (0.0227)
Observations 2,720 2,720 2,720 2,720
Table 6: Effect of Photos on Evasion – Hurdle Model

Model
(1) (2) (3) (4)
Explanatory Variable Probit Tobit Probit Tobit Probit Tobit Probit Tobit
Photos -1.256*** 0.222*** -1.307*** 0.225*** -1.865* 0.369*** -0.890*** 0.287***
(0.272) (0.037) (0.299) (0.0366) (1.075) (0.057) (0.165) (0.0512)
Low Audit 0.105 0.146*** -0.586 0.263*** 0.105 0.210***
(0.0864) (0.0303) (1.025) (0.054) (0.176) (0.0333)
Photos * Low Audit 0.79 -0.255*** 0.0982 -0.204***
(1.031) (0.0725) (0.195) (0.0543)
Italy 4.275 -0.259***
(92.84) (0.0372)
Italy * Photos -3.92 0.0358
(92.84) (0.0588)
Low Penalty 0.00593 0.189***
(0.0748) (0.0262)
Constant 1.323*** 0.242*** 1.326*** 0.161*** 1.819* 0.0987* 0.677*** 0.207***
(0.291) (0.0394) (0.333) (0.0455) (1.091) (0.0543) (0.166) (0.0425)
Sigma 0.496*** 0.493*** 0.489*** 0.455***
(0.0186) (0.0185) (0.018) (0.0122)
Observations 2,720 2,720 2,720 2,720 2,720 2,720 2,720 2,720
30

103.4.95.4 on Mon, 17 Jul 2023 05:02:30 +00:00
Table 7: Effect of Pictures on Evasion – Hurdle Model 4 at Various Levels of Being Caught
Model
(1) Ever Caught (2) Times Caught>1 (3) Times Caught>2 (4) Times Caught>3
Explanatory Variable Probit Tobit Probit Tobit Probit Tobit Probit Tobit
Photos -0.581*** 0.265*** -0.452 0.307*** 0.449 0.272** 1.312* 0.0971
(0.224) (0.0595) (0.281) (0.0866) (0.371) (0.137) (0.691) (0.207)
Low Audit 0.102 0.240*** 0.265 0.237*** 5.513 0.0999 0.383 0.121
(0.231) (0.0400) (0.286) (0.0550) (272.8) (0.0836) (0.523) (0.164)
Photos * Low Audit 0.232 -0.166** 0.0794 -0.186* -7.864 0.234 -1.162 0.0699
(0.275) (0.0653) (0.368) (0.0974) (272.8) (0.147) (0.803) (0.259)
Italy 4.349 -0.226*** 5.132 -0.200*** 5.652 -0.104 5.409 -0.345**
(142.2) (0.0433) (811.7) (0.0601) (298.6) (0.0855) (186.2) (0.142)
Italy * Photos -3.992 0.105 -4.823 0.106 -4.586 -0.187 -5.746 0.284
(142.2) (0.0665) (811.7) (0.0944) (298.6) (0.145) (186.2) (0.225)
Low Penalty 0.0540 0.161*** 0.00986 0.124*** 2.278* 0.0579 0.446 0.112
(0.121) (0.0332) (0.173) (0.0477) (0.692) (0.0745) (0.492) (0.165)
**
Constant 0.730*** 0.243*** 0.572** 0.254*** -0.218 0.296*** -0.238 0.503***
(0.231) (0.0512) (0.275) (0.0703) (0.294) (0.0930) (0.382) (0.143)
Sigma 0.424*** 0.433*** 0.462*** 0.402***
(0.0137) (0.0191) (0.0279) (0.0407)
Observations 1,337 1,337 728 728 312 312 113 113
31

103.4.95.4 on Mon, 17 Jul 2023 05:02:30 +00:00
Table 8: Effect of Photos on Evasion – Random Effects Hurdle Model
Model 1 Model 2 Model 3 Model 4
Explanatory Variable Hurdle Above Hurdle Above Hurdle Above Hurdle Above
Photos - -0.0435 - -0.0453 - 0.0138 - 0.108
(0.286) (0.0810) (0.285) (0.0801) (0.285) (0.0833) (0.296) (0.0843)
0.968*** 0.968*** 0.967*** 0.919***
Low Audit 0.163*** 0.219*** 0.221***
(0.0218) (0.0305) (0.0299)
Photos * Low Audit - -
(0.0434) (0.0425)
0.114*** 0.118***
Italy -0.114
(0.0801)
Italy * Photos 0.0096
(0.113)
Low Penalty 0.178***
(0.0213)
Constant 1.642*** 0.195*** 1.641*** 0.114*** 1.640*** 0.0854* 1.584*** 0.0777
(0.248) (0.0425) (0.248) (0.0439) (0.247) (0.0453) (0.256) (0.0601)
Sigmau 0.309*** 0.309*** 0.309*** 0.331***
(0.0261) (0.0262) (0.0262) (0.0278)
Sigmae 0.487*** 0.480*** 0.478*** 0.468***
(0.0102) (0.0100) (0.01000) (0.00977)
Transformed Rho 0.233 0.247 0.245 -0.533
(0.473) (0.474) (0.475) (0.473)
Observations 2,720 2,720 2,720 2,720 2,720 2,720 2,720 2,720
Standard errors are in parentheses, *** p<0.01, ** p<0.05, * p<0.1
32

103.4.95.4 on Mon, 17 Jul 2023 05:02:30 +00:00
Figure 1: Tax Filing Decision
33

103.4.95.4 on Mon, 17 Jul 2023 05:02:30 +00:00
Figure 2: Audit and Post-Audit Information
34

103.4.95.4 on Mon, 17 Jul 2023 05:02:30 +00:00
Figure 3: CDF of Evasion Rates by Treatment
35

103.4.95.4 on Mon, 17 Jul 2023 05:02:30 +00:00
APPENDIX A: EXPERIMENT INSTRUCTIONS
Instructions (No Information)
This is an experiment about economic decision making. The study will last no more than 2 hours.
You will receive $10 for your participation and will have the opportunity to increase this amount
based on the decisions you make. Your earnings will be paid to you in cash at the end of the
study. Your decisions and payments will be kept private. How your decisions affect your
earnings is explained below.
The decisions made in this experiment are tax reporting decisions. In each round, you will be
given an income of $50. You will have to report your income to a tax authority and pay taxes on
reported income. The tax rate is 35%. Thus, your taxes will be 0.35* (reported income). After
you submit your taxes, there is a chance that you will be audited by the tax authority. The
probability of audit is 30%, and the computer will use a random number generator to decide
whether the audit will occur. If you are audited and you have reported less than your full income,
this will be detected by the audit and you will be required to pay a fine on the unpaid taxes. You
will pay two times the unpaid taxes. In other words, if you have unreported income and you are
audited, a penalty which equals 2.0*0.35*(actual income - reported income) will be subtracted
from your after-tax-income to get your final income for the round.
How your earnings will be determined

If you are not audited, your earnings for the round will be
Actual income – 0.35* (reported income)
If you are audited, your earnings for the round will be

Actual income-taxes paid-penalty for undisclosed income
=actual income-0.35*(reported income)-2*0.35 (actual income-reported income)
Examples
These examples will demonstrate the type of decision you will be making and how your earnings
will be determined.
Example 1. Suppose your income for the round is $50 and that you report $50 as your income.
Then you will pay 0.35*$50.00 = $17.50 in taxes.
If not audited, your earnings for the round will be $50 - $17.50 = $32.50
If audited, your earnings for the round will be $50 - $17.50 = $32.50
Example 2. Suppose your income for the round is $50.0 and that you report $30.00 as your
income. Then you will pay 0.35*$30 = $10.50 in taxes
If audited, the audit will detect unreported income of $20.00 ($50 - $30). You will be
required to pay 2 times the tax owed on this unreported income which equals
2*0.35*($20) = $14. Your earnings for the round will be $50 - $10.50 - $14 =
25.50
The experiment will have 10 rounds, but only one of these will count for payment. At the end of
the experiment, a 10-sided die will be rolled to determine which round will count for payment.
After the round is selected for payment, you will be paid in cash your earnings in that round.
Each round is equally likely to be selected, but you will not know in advance which one will be
chosen.
36

103.4.95.4 on Mon, 17 Jul 2023 05:02:30 +00:00
If you have any questions, please raise your hand and one of us will come to your desk to answer
it.
Instructions (Information)
This is an experiment about economic decision making. The study will last no more than 2 hours.
You will receive $10 for your participation and will have the opportunity to increase this amount
based on the decisions you make. Your earnings will be paid to you in cash at the end of the
study. Your decisions and payments will be kept private. How your decisions affect your
earnings is explained below.
The decisions made in this experiment are tax reporting decisions. In each round, you will be
given an income of $50. You will have to report your income to a tax authority and pay taxes on
reported income. The tax rate is 35%. Thus, your taxes will be 0.35* (reported income). After
you submit your taxes, there is a chance that you will be audited by the tax authority. The
probability of audit is 30%, and the computer will use a random number generator to decide
whether the audit will occur. If you are audited and you have reported less than your full income,
this will be detected by the audit and you will be required to pay a fine on the unpaid taxes. You
will pay two times the unpaid taxes. In other words, if you have unreported income and you are
audited, a penalty which equals 2.0*0.35*(actual income - reported income) will be subtracted
from your after-tax-income to get your final income for the round.
How your earnings will be determined

If you are not audited, your earnings for the round will be
Actual income – 0.35* (reported income)
If you are audited, your earnings for the round will be

Actual income-taxes paid-penalty for undisclosed income
=actual income-0.35*(reported income)-2*0.35 (actual income-reported income)
37

103.4.95.4 on Mon, 17 Jul 2023 05:02:30 +00:00
Examples
These examples will demonstrate the type of decision you will be making and how your earnings
will be determined.
Example 1. Suppose your income for the round is $50 and that you report $50 as your income.
Then you will pay 0.35*$50.00 = $17.50 in taxes.
If audited, your earnings for the round will be $50 - $17.50 = $32.50
Example 2. Suppose your income for the round is $50.0 and that you report $30.00 as your
income. Then you will pay 0.35*$30 = $10.50 in taxes
If audited, the audit will detect unreported income of $20.00 ($50 - $30). You will be
required to pay 2 times the tax owed on this unreported income which equals
2*0.35*($20) = $14. Your earnings for the round will be $50 - $10.50 - $14 =
25.50
The experiment will have 10 rounds, but only one of these will count for payment. At the end of
the experiment, a 10-sided die will be rolled to determine which round will count for payment.
After the round is selected for payment, you will be paid in cash your earnings in that round.
Each round is equally likely to be selected, but you will not know in advance which one will be
chosen.
In each round, you will be shown the income reported by 25 percent of the subjects in this
experiment. After you have viewed your earnings information from the current round, you will
be shown a table showing the level of reported income by 25 percent of the people in this
session. You will see a 3-digit ID code of a subject and the income reported by this subject. You
will NOT be told which ID code belongs to which person in this room.
Every round 25 percent of your returns will be randomly chosen to be shown to the others, thus
the returns you see may be for different subjects in different rounds.
If you have any questions, please raise your hand and one of us will come to your desk to answer
it.
38

103.4.95.4 on Mon, 17 Jul 2023 05:02:30 +00:00

Culture Compliance and Confidentiality Study of Taxpayer Behavior in The United States

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Culture Compliance and Confidentiality Study of Taxpayer Behavior in The United States

Uploaded by

Copyright:

Available Formats

National Tax Association

Culture, Compliance, and Confidentiality

Stable URL: https://www.jstor.org/stable/10.2307/26816608

This content downloaded from

Culture, Compliance, and Confidentiality:

JEL Classifications: H2, H3.

Keywords: Tax compliance, experimental economics, confidentiality, social norm.

This content downloaded from

compliance. Traditional compliance-inducing measures include penalty rates and audit

compliance. Therefore, whether public disclosure of taxpayer compliance behavior increases or

the impact of confidentiality of taxpayer information on the level of individual compliance in

This content downloaded from

Several laboratory experiments have looked at the effects of disclosure on taxpayer

cheating and disclosure in a tax-reporting experiment. In a “pictures” treatment, a subject who

This content downloaded from

emotional arousal) after an audit.

Despite these contributions, the impact of confidentiality versus disclosure of taxpayer

Confidentiality”), an individual who is detected evading is financially penalized, but their

with information on the level of underreporting).

This content downloaded from

stigma influences reporting decisions.

marginal impact of evasion disclosure on compliance behavior than would be possible by

focusing on just one country.

disclosure encourages compliance.

2. The Institutional Context

and toward tax compliance – in these two countries.

This content downloaded from

political office choose to do so.

This content downloaded from

impact in reducing evasion.

This content downloaded from

available in the web.

This content downloaded from

Following Allingham and Sandmo (1972), tax compliance is typically modeled as a

income IC if caught underreporting equals

underreporting is not caught, income IN is

This content downloaded from

income to maximize expected utility, defined as

(3) EU(I)=pU(IC )+(1-p)U(IN ),

p and the penalty rate f unambiguously increase declared income.

compliance by increasing the audit and the penalty rates.

However, it is clear to many observers that compliance cannot be explained entirely by

This content downloaded from

the financial incentives they face from the enforcement regime.

ways not captured by the basic economics-of-crime approach. Motivations affected by

confidentiality may well play a role here.

Indeed, it is possible to extend the basic economics-of-crime approach to incorporate the

distinguished by the feature that it is process-oriented, unlike the outcome-orientation of

disapproval. Consequently, if others behave according to some socially accepted mode of

appeals to patriotism, conscience, or upon an individual’s feelings of altruism, fairness, morality,

This content downloaded from

compliance and, through this channel, each individual’s compliance decision.

their full taxes will suffer a loss in utility.

than in the basic economics-of-crime model discussed earlier.

Clearly, γ is likely to be sensitive to the disclosure of one’s compliance behavior to others;

This content downloaded from

enhance the effect of disclosure.

Of course, disclosure could instead lead to a decline in γ if noncompliance is viewed as a

compliance as a statement against government.

This content downloaded from

underreported income may reduce the overall compliance rate.

This discussion illustrates the complexity of the compliance effects of breaking

confidentiality. As emphasized earlier, the actual evidence on confidentiality versus disclosure is